Speed up bash script while loop-CodePudding

This script redirects the output of a hexdump to a text file and then loops through it printing out rows. There are exactly 374371 rows of data in the file that look like this:

1a 03 1a 03 4a 03 57 03
4b 03 44 03 1e 03 09 04
10 03 19 03 40 03 ae 03
1e 03 26 03 33 03 ad 03
10 03 84 03 43 03 62 03

Here is the script:

if [ ! -d ./temp ]; then
  mkdir ./temp
  mount -t tmpfs -o size=128m tmpfs ./temp

  if [[ -s ./temp/samples.txt || ! -f ./temp/samples.txt ]]; then
    hexdump -e '8/1 "x " "\n"' samples.bin > ./temp/samples.txt
    echo "              ch0    ch1    ch2    ch3"
    x=0
    while read line;do
      echo "Sample "$x":        "$(echo $line | awk '{print "0x"$1$2"  0x"$3$4"  0x"$5$6"  0x"$7$8}')
      x=$((x 1))
    done<./temp/samples.txt
  fi
  umount ./temp
  rm -Rf ./temp
fi

The output looks like:

                ch0    ch1    ch2    ch3
Sample 0:       0x1a03 0x1a03 0x4a03 0x5703
Sample 1:       0x4b03 0x4403 0x1e03 0x0904
Sample 2:       0x1003 0x1903 0x4003 0xae03
Sample 3:       0x1e03 0x2603 0x3303 0xad03
Sample 4:       0x1003 0x8403 0x4303 0x6203
Sample 5:       0xe003 0x1603 0x3403 0xc403
Sample 6:       0xf802 0x3b03 0x5303 0x6103
Sample 7:       0x1003 0x1503 0x4203 0x5803
Sample 8:       0x2303 0x1f03 0x5703 0x6203
Sample 9:       0x1703 0x7303 0x3103 0x3303
Sample 10:      0x1403 0xff02 0x3003 0x5103
Sample 11:      0x5f03 0x4203 0x4703 0x7e03
Sample 12:      0xba03 0x2603 0x3503 0xa003

Even with tmpfs, the script doesn't run any faster than just reading from disk. Is there a way to make this script run any faster?

CodePudding user response：

If you want to speed-up the script, consider to use awk instead of bash as Shellter comments. Besides you may don't need to use hexdump. Just concatenate every two nibbles. Then would you please try:

awk '
    BEGIN {
        print "                ch0    ch1    ch2    ch3"        # print the header line
    }
    {
        printf("Sample %d:", NR - 1)                            # print the sample number
        printf(substr("        ", 1, 8 - length(NR - 1)))       # adjust the length of the spaces
        for (i = 1; i <= NF; i =2) {                            # print every two nibbles
            j = i   1
            printf("0x%s%s%s", $i, $j, j == NF ? ORS : OFS)
        }
    }
' samples.txt

Output with the provided file:

                ch0    ch1    ch2    ch3
Sample 0:       0x1a03 0x1a03 0x4a03 0x5703
Sample 1:       0x4b03 0x4403 0x1e03 0x0904
Sample 2:       0x1003 0x1903 0x4003 0xae03
Sample 3:       0x1e03 0x2603 0x3303 0xad03
Sample 4:       0x1003 0x8403 0x4303 0x6203

CodePudding user response：

With your shown samples and attempts please try following awk code. Written and tested in GNU awk. Reading Input_file 2 times but first time it reads only very first line to get TOTAL number of fields and moves to the 2nd time its run.

awk '
BEGIN{
  print"\tch0    ch1    ch2    ch3"
}
FNR==NR{
  totalCount=NF
  nextfile
}
RT{
  gsub(/^[[:space:]]|\n $/,"",RT)
  sub(/ /,"",RT)
  val=(val?val OFS:"") "0x1" RT
  if(  count%(totalCount/2)==0){
     print "Sample "  count2":\t"val
     val=count=""
  }
}
' Input_file RS="(^|\n| )[[:alnum:]]{2} [[:alnum:]]{2}" Input_file | 
column  -t -s $'\t'

CodePudding user response：

It is also possible with sed. I found sed in general to be faster than awk.

cat -n | sed 's/\s*\([0-9]*\)\s*\([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\) \([0-9a-f]*\)/Sample \1:\t0x\2\3 0x\4\5 0x\6\7 0x\8\9/'